WIP: Attempting to test against conda-build master and the latest released version of conda-build

[ericdill]

I could have done this by turning on travis for my fork, but this is easier.
I'm also doing this because I think I already found an API change in conda-build
master that breaks resolved_distribution.py. Let's see if the test suite on conda-build-all hits it.
If not, I'll make a test to expose it

[pelson]

Thanks @ericdill - interested to see what has changed with conda-build. Definitely don't want you to drop this investigation.

[ericdill]

@pelson At the end of this circle-ci build is what I think the problem is. I dug a little into the source for conda-build-all and this self.meta.parse_again() call gained a positional argument in conda-build master. https://github.com/conda/conda-build/blob/33135a88a103c566536a7fc40868df73d2795213/conda_build/metadata.py#L330

[pelson]

Thanks @ericdill. conda/conda-build#742

[ericdill]

Oh, great. Glad you found the problem and already fixed it! Does it still make sense to try to test against current master of conda-build?

[pelson]

Does it still make sense to try to test against current master of conda-build?

That's a hard call. In some ways yes, in otherways, we don't test all master versions of dependencies (e.g. python, conda, etc.).

[ericdill]

Yeah, I suppose its better to just fix upstream libraries when bugs are discovered. No reason to make the automated testing more opaque than it needs to be.